Time Nick Message 00:11 paramat crash when using 'decrease view range' #6400 00:11 ShadowBot https://github.com/minetest/minetest/issues/6400 -- Crash 'buffer overflow detected' when using decrease view range 00:13 Fixer paramat: on newest does not crash 00:13 paramat ok will try latest 00:14 paramat although i can't see a commit that would fix 00:15 rubenwardy #6397 00:15 ShadowBot https://github.com/minetest/minetest/issues/6397 -- Add warning on attempt to revoke irrevokable privileges by rubenwardy 00:23 paramat rubenwardy can i merge #6284 now? 00:23 ShadowBot https://github.com/minetest/minetest/issues/6284 -- Fix empty legacy meta being persisted by rubenwardy 00:23 paramat with #6385 00:23 ShadowBot https://github.com/minetest/minetest/issues/6385 -- Schematic decorations: Add 'place_offset_y' placement parameter by paramat 00:23 rubenwardy yeah sure 00:23 paramat will do 00:27 paramat will merge in a moment 00:36 paramat done 00:51 paramat crash (issue 6400) still happens on latest master, clean build, weird 00:52 paramat aha 00:52 paramat saw your comment 00:52 rubenwardy I'm just bisecting now 00:52 paramat in a way i'm relieved 00:53 paramat i wonder if related to #6390 00:53 ShadowBot https://github.com/minetest/minetest/issues/6390 -- Occasional crash in recent commit 40dd03e 00:55 rubenwardy paramat: https://github.com/minetest/minetest/pulse 00:55 paramat i seem to remember it did not happen in commit a3c298e 00:55 rubenwardy it's looking good 00:55 rubenwardy shame it doesn't show closed PRs 00:56 rubenwardy the month range is even better 00:56 paramat nice 00:57 paramat lol 161 closed issues! 00:58 rubenwardy you have twice as many commits as me in that period 00:58 * rubenwardy needs to catch up 01:00 rubenwardy huh, we didn't have a dev meeting today 01:02 paramat or last week. shadowninja is busy. maybe sundays are better 01:10 rubenwardy paramat, done 01:10 rubenwardy it was Wuzzy, it was 01:12 paramat aah 01:13 paramat thanks 01:13 rubenwardy building with debug now 01:14 rubenwardy #6400 01:14 ShadowBot https://github.com/minetest/minetest/issues/6400 -- Crash 'buffer overflow detected' when changing view range 01:16 rubenwardy almost certainly to do with swprintf 01:17 rubenwardy wth 01:17 rubenwardy I can't reproduce it now 01:19 rubenwardy any comment on #6397, paramat 01:19 ShadowBot https://github.com/minetest/minetest/issues/6397 -- Add warning on attempt to revoke irrevokable privileges by rubenwardy 01:19 rubenwardy ~? 01:21 paramat had a quick look but need to review more carefully, will do so later 01:22 rubenwardy ha, the crash only happens on release builds 01:22 rubenwardy yay! 01:23 paramat also crashes on F6 press 01:24 rubenwardy for integers, it's better to use a string stream instead of sfprint 01:24 rubenwardy and I'm right, it's sfprint related 01:24 rubenwardy *swprintf 01:29 paramat can't seem to show minimap either 01:29 paramat unless i'm doing something silly 01:31 paramat no ignore that 01:31 paramat i disabled that in conf :P 01:36 rubenwardy found the issue 01:36 rubenwardy he's using sizeof(buf) instead of sizeof(buf) / sizeof(wchar_t) 01:37 rubenwardy also, we really should look into using std::wstring and a formating function for it 01:37 rubenwardy we should be using buffers in game.cpp 01:37 rubenwardy *shouldn't 01:41 rubenwardy paramat, https://github.com/rubenwardy/minetest/commit/d01d246d5afcae6d50fcb1fd61c4526affc943f7 01:41 rubenwardy would you agree that's trivial? 01:41 rubenwardy yeah, it's trivial 01:41 rubenwardy I'll merge in 10 01:42 paramat looking 01:42 rubenwardy basically, sizeof gets the total size of a thing in bytes 01:42 paramat fine 01:43 rubenwardy it's a common mistake, tbh 01:52 rubenwardy meging 01:54 rubenwardy done. easy mistake to make 01:55 rubenwardy Makes me question why it wasn't picked up in review though 01:55 paramat good, thanks 01:56 rubenwardy that's probably the quickest closed bug issue ever 02:01 rubenwardy only 30 commits until we reach 7000 02:14 paramat ugh #1566 02:14 ShadowBot https://github.com/minetest/minetest/issues/1566 -- saving username/password login per server basis 02:20 rubenwardy #6401 02:20 ShadowBot https://github.com/minetest/minetest/issues/6401 -- Add check to pause game on lost window focus by rubenwardy 02:20 rubenwardy sorry paramat 02:22 rubenwardy note that you can open the inventory to still use continous forward 02:30 paramat no prob. yeah i use inventory that way 02:36 rubenwardy anyway, that's enough for today 02:36 rubenwardy i/ 02:36 rubenwardy * o/ 02:41 paramat 0/ 06:05 nore Someone broke lint :/ 14:07 rubenwardy grrr 14:07 rubenwardy why does the lint complain about code in a completely different file to what I've touched? 14:31 nore rubenwardy: I think it was broken in a PR that was merged without noticing it 14:31 nore I got the same problem in my pr 14:39 nore what are the rules about one's own approval again? (for #6398) 14:39 ShadowBot https://github.com/minetest/minetest/issues/6398 -- Cleanup & document better conversions between block position and integers by Ekdohibs 14:41 rubenwardy is it possible that that would cause a different calculation of a pos hash? 14:42 rubenwardy you mention it fixes a bug, but that bug fix could cause problems if the map format relies on it 14:42 sfan5 rubenwardy: as the text explains, no 14:42 sfan5 but i still feel unwell about this, someone should actually test this 14:43 sfan5 nore: self-approvals are possible 14:43 nore sfan5: I tested it, didn't break my map 14:43 sfan5 that's not the kind of testing i mean 14:43 nore also, the function was buggy but that bug didn't matter, and I got rid of it anyway 14:43 nore ah 14:44 nore the fact that it does the same thing? 14:44 sfan5 *if* this pr would cause a compat break it would only manifest in some edge cases 14:44 sfan5 yes 14:45 sfan5 while i agree with your explanation, imo a test should be done with a wide range of values to ensure the function still performs as expected 14:45 sfan5 (why don't we have unittests for this anyway?) 14:45 nore no idea 14:46 sfan5 maybe it's time to add some 14:48 nore hmm, yeah 14:49 nore anyway, just a quick check gave this: https://gist.github.com/Ekdohibs/870a2af8950f3147f4d912f59909d897 14:50 sfan5 since this is python you wouldn't need pmod at all 14:50 sfan5 but if that works then it's fine by me 14:57 rubenwardy updated #6396 14:57 ShadowBot https://github.com/minetest/minetest/issues/6396 -- Add reasons to on_dieplayer and on_hpchange by rubenwardy 14:58 rubenwardy now just need to let mods provide reasons 15:15 Krock also, why do we use such a strange blockToInteger conversion? three s16 would fit perfectly into this s64 memory, using bitops or memcpy 15:17 nerzhul Krock, it's historical, maybe git blame can tell you why 15:18 nerzhul you talk about the position split in sqlite, right ? 15:18 Krock rather position merge than split, as they're squashed into a s64 variable 15:21 Krock masking was already used before 41bc244, but seemed to be a incomplete improvement 15:23 Krock and implemented in 58841ef, with the other database backends 15:23 nerzhul note, it's eaxct it's faster with merged value than 3 values 15:28 Krock yes, they are. Found out that the function was only moved in that commit. First written in Oct 2011 (v 0.4.6) 15:29 Krock wrong tag. * 0.3.0 15:33 nore sfan5: I would need pmod, since the whole point of this pr was to get rid of that pythonmodulo function that didn't do as adversised 15:33 nore ((-4096)%4096 is 0 in python, not 4096) 15:34 nerzhul don't trust python, some maths are weird 15:34 nerzhul especially in modulo 15:34 sfan5 oh right 15:34 nore nerzhul: actually python modulo is consistent, as opposed to C modulo 15:35 nerzhul do you know why ? because python does maths, C does memory operations :) 15:35 nore python modulo is always in [[0, mod-1]], while C modulo depends on the sign of the argument 15:35 nore exactly :) 15:35 nore but in that case it doesn't matter anyway since we only give nonnegative arguments to it 15:48 celeron55_ i don't know what's happening but ShadowNinja or someone like that already broke the conversion at some point and messed up some people's worlds 15:48 celeron55_ and it was reverted 15:48 celeron55_ don't do the same mistake again 15:49 celeron55_ i guess the short story is, it doesn't do what most C programmers think it does 15:51 nore celeron55_: probably, I made sure it is correct 15:51 nore if you just do stupid masking it doesn't work 15:56 nore https://github.com/Ekdohibs/minetest/commit/54ffe2e5de9ce44129f84f4748743f893b75fda7#diff-6ef17526be0e4cc9b0f882af87c18a4aR44 <-- the cause was that it was using shifting instead of find the value, removing it then masking again 16:07 celeron55_ ok looked at the PR you were talking about now and it looks like you improved upon whatever SN had left 16:11 celeron55_ it might be beneficial to add a note like "this is NOT a bitmask" 16:12 celeron55_ because SN tried to interpret it as a bitmask and i think someone else might do the same unless told not to 16:12 nerzhul celeron55_, a version in db should be added + migration if done, yes 16:14 celeron55_ the conversion came from the first sqlite implementation contributed by a server admin sometime before 0.3 16:14 celeron55_ if that answers Krock's question 16:15 celeron55_ that's all that is known about the reasoning anyway, i.e. not much 16:24 Krock yes, thanks celeron55_ 18:21 Krock Quick patch to fix the selectionbox being erased if no collisionbox is set: https://pastebin.com/raw/70eSBjdK 18:23 Krock Will merge in ~10 minutes if there are no objections 18:24 rubenwardy makes sense 21:27 Fixer not decaying trees bug found on Calinou's server: https://i.imgur.com/brYMtss.png 21:33 Calinou heh 21:34 Calinou the one to the left looks like an upvote 23:52 Fixer please announce on forum about nodeupdate removal, a lot of mods still use this, modders should be warned 23:59 Fixer quick search on github: https://github.com/search?p=1&q=filename%3A%22.lua%22+nodeupdate&type=Code&utf8=%E2%9C%93 23:59 Fixer my local archive search: https://pastebin.com/raw/namAt0aM